US 11,727,289 B2
Iterative generation of top quality plans in automated plan generation for artificial intelligence applications and the like
Michael Katz, Elmsford, NY (US); Shirin Sohrabi Araghi, Port Chester, NY (US); and Octavian Udrea, Valhalla, NY (US)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on May 4, 2018, as Appl. No. 15/971,911.
Prior Publication US 2019/0340525 A1, Nov. 7, 2019
Int. Cl. G06N 5/045 (2023.01); G06F 8/30 (2018.01); G05B 19/042 (2006.01); G05B 13/04 (2006.01)
CPC G06N 5/045 (2013.01) [G05B 13/041 (2013.01); G05B 19/0426 (2013.01); G06F 8/31 (2013.01)] 13 Claims
OG exemplary drawing
 
1. A method for improving performance of at least one hardware processor solving a top-k planning problem, said method comprising:
obtaining, in a memory coupled to the at least one processor, a specification of the planning problem in a planning language;
obtaining, in a first iteration carried out by said at least one processor, at least one solution to said planning problem;
modifying, in said first iteration carried out by said at least one processor, said planning problem to forbid said at least one solution; and
repeating, by said at least one processor, said obtaining of said at least one solution and said modifying to forbid said at least one solution, for a plurality of additional iterations, after said first iteration, until a desired number, k, of solutions to said planning problem are found or until no further solutions exist, whichever comes first;
wherein:
in said step of obtaining said at least one solution in said first iteration and said plurality of additional iterations, said at least one solution comprises an optimal solution; and
said step of modifying, in said first iteration and said plurality of additional iterations, comprises modifying to forbid said optimal solution;
at least one forbidden solution of the at least one solution comprises a different set of actions than an unforbidden solution of the at least one solution;
further comprising, for said first iteration and each of said additional iterations, extending said obtained optimal solution to an extended set of solutions including said optimal solution, wherein said modifying of said planning problem to forbid comprises modifying said planning problem to forbid said extended set of solutions;
wherein, in said step of obtaining said specification of said planning problem in said planning language, said specification specifies a problem in automated control of an industrial robot, further comprising operating said industrial robot in accordance with said k solutions to said planning problem and wherein said extending of said obtained optimal solution to said extended set of solutions comprises adding to said extended set of solutions new set members symmetric to already existing set members, wherein said set members symmetric to said already existing set members comprise set members resulting from mapping said existing set members with structural symmetries.